This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Your last commit to new-regalloc-branch
- To: Daniel Berlin <dan at cgsoftware dot com>
- Subject: Re: Your last commit to new-regalloc-branch
- From: Denis Chertykov <denisc at overta dot ru>
- Date: 05 Aug 2001 11:38:17 +0400
- Cc: Denis Chertykov <denisc at overta dot ru>, gcc-patches at gcc dot gnu dot org, matzmich at cs dot tu-berlin dot de
- References: <auto-000000404396@overta.ru> <8766c3twbs.fsf@cgsoftware.com>
Daniel Berlin <dan@cgsoftware.com> writes:
> If this works, it means we need to fix regclass and regscan_update to
> just frickin work.
> We can't go redoing all of life analysis, all of register scanning,
> all of register usage, etc, on each pass.
> It's murder.
I'm agree with you. My patch significantly slowdown the new-allocator.
I don't apply my patch.
> reg_scan_update was specifically there for incremental updating when
> you emit new pseudos.
> If it doesn't work, we should fix it.
> max_regno was also correct given the circumstances. reg_scan_update
> wanted the previous max register (IE the max before we emitted new
> pseudos), so max_regno, having not been updated, was that number.
Yes yes.
Denis.